The Experiment Leaderboard is the Experiments view that ranks your deployed control and variant pages side by side by their observed outcomes, so you can see at a glance which version is performing best across the pages under test.
| Term | Experiment Leaderboard |
|---|---|
| Category | AI, Content Generation and Experiments |
| Also known as | Leaderboard |
| Where it appears | Experiments |
What it means in RankGear
The Leaderboard lives inside Experiments, where you run a controlled test of a control page against one or more variants. It is the scoreboard for that test: each deployed page appears as a row, paired with the outcome RankGear has recorded since it went live, and the rows are ordered so the strongest performer sits at the top. What you are reading is measured results from pages that are actually deployed — not a forecast or a target, but where each version stands against the others right now.
How to interpret it
Read the order as a comparison, not a verdict. The ranking reflects outcomes gathered over the experiment window, so a short run or thin traffic can leave the positions unstable and reshuffle them as more data arrives. Let the test accumulate enough time and volume before you treat the top row as the winner. A variant sitting in first place is evidence that it is out-performing its rivals in the measured window — it is not proof that the change caused the difference. Confirm the pages differ only by the thing you set out to test, because without that isolation the gap could come from something you never intended to vary.
| What a row shows | What it tells you |
|---|---|
| Page (control or variant) | Which deployed version this outcome belongs to, so you can tell the baseline from each alternative. |
| Observed outcome | The result RankGear has measured for that page over the experiment window — the value the ranking is sorted on. |
| Position | Where this page stands relative to the others; the leader is the current front-runner, not a settled result. |
Example
Say you are testing the intro on a category page. The control keeps the original 120-word opener, variant A leads with a single answer-first paragraph, and variant B adds an FAQ block beneath it. A few weeks after all three are deployed, the Experiment Leaderboard ranks variant A first on the outcome you chose to track, the control second, and variant B third — telling you the answer-first rewrite is the front-runner while the FAQ variant is trailing the page it was meant to beat. That order points you toward variant A, but you would still let the window run and confirm the three pages differ only by the intro before promoting it.
Important considerations
- Observed, not projected: the ranking is built from outcomes measured over the experiment window, so it needs enough time and traffic to settle. Early positions can swing before the data stabilizes.
- Correlation is not causation: a variant on top is out-performing its rivals in the measured window, not proven to have caused the result. Verify experimental isolation — that the pages differ only by the change under test.
- Check AI-assisted variants: if a variant’s content or an experiment critique was drafted with AI, the output can be incomplete or wrong. Verify factual claims, statistics, and citations before you trust or publish it.
- Comparative indicator, not a Google score: the outcomes are signals RankGear records across the pages under test, not values Google reports, and a page winning the leaderboard does not by itself make it rank.
Related terms
Part of the RankGear glossary · how RankGear measures · the 870 factors.